Methods to adjust for misclassification in the quantiles for the generalized linear model with measurement error in...

Area of Science:

• Biostatistics
• Epidemiology
• Health Research Methods

Background:

• Biomedical studies often use continuous exposure variables as covariates in regression analyses.
• Measurement errors in these exposure variables can lead to misclassification of exposure quantiles.
• This misclassification can introduce bias into the estimated association between exposure and outcome, particularly when gold standard data are unavailable.

Purpose of the Study:

• To develop and evaluate novel statistical methods for adjusting bias in effect estimation caused by measurement error in exposure quantiles.
• To provide practical and robust estimators for regression analyses involving misclassified exposure variables.

Main Methods:

• Development of two regression calibration estimators: a normal likelihood-based estimator and a linearization-based estimator.
• Evaluation of estimator performance through a simulation study to assess finite sample properties.
• Application of the developed methods to real-world data from a randomized clinical trial on exercise and weight loss interventions.

Main Results:

• The proposed regression calibration estimators effectively reduce bias in effect estimation arising from exposure quantile misclassification.
• The linearization-based estimator offers a simple and practical approach for bias adjustment.
• Simulation results demonstrate the utility and accuracy of the proposed methods under various scenarios.

Conclusions:

• Measurement error in continuous exposure variables and subsequent quantile misclassification pose significant challenges in biomedical research.
• The developed regression calibration techniques provide effective solutions for mitigating bias in association estimates.
• These methods enhance the reliability of findings from observational and interventional studies where exposure measurement error is present.